Measuring the value of housing services in household surveys: an application of machine learning approach
|
|
- Beryl Gibson
- 6 years ago
- Views:
Transcription
1 Measuring the value of housing services in household surveys: an application of machine learning approach Weldensie T. Embaye and Yacob A. Zereyesus* *Department of Agricultural Economics, Kansas State University, Manhattan, KS, USA Contact Author: Selected Paper prepared for presentation at the Southern Agricultural Economics Association (SAEA) Annual Meeting, Mobile, Alabama, February 4-7, Copyright 2017 by Weldensie T. Embaye and Yacob A. Zereyesus. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies. 1
2 Abstract Housing value is a major component of the aggregate expenditure used in the analyses of welfare status of households in the development literature. Accurate measurement of housing services is therefore important to obtain the value of housing in household surveys. Data shows that a significant proportion of households in a typical household survey adopted by the Word Bank are self-owned. The standard approach of imputing the housing value for such surveys is based on what it would cost to rent the house. This hedonic pricing is normally predicted using an Ordinary Least Squares (OLS) method. Literature shows that machine learning methods have better predictive power over OLS applied in other valuation exercises. We tested whether or not several machine learning methods (e.g. Ridge, Lasso, Tree, Random forest and Boost) provide superior prediction values of housing over OLS using cross country household survey data. In general, our results confirm that machine learning methods have better predictive capacity than OLS over the given data sets. While predicting housing values, recent developments in the machine learning approaches adopted in the study could be used to obtain improved predictive outcomes. Keywords: Housing values, machine learning, OLS, cross country evidence 2
3 Introduction The sustainable reduction of poverty and hunger are among the Millennium Development Goals (MDG) as adopted by the World Bank and world community. According World Bank (2012), world poverty had declined from 1.94 billion in 1981 to 1.29 in To better understand the dynamics of poverty and to design appropriate and effective poverty reduction policies, it is imperative to have well documented poverty data at household level. Many NGOs in collaboration with the national statistics offices have been participating in household level data collection. The World Banks s Living Standards Measurement Study-Integrated Surveys on Agriculture (LSMS- ISA) and others such as the Bill and Melina Gates Foundation cooperate with the national statistics offices on collecting panel based household surveys on many countries counties in Sub-Saharan Africa. Household aggregate expenditure is frequently used for welfare analyses. Housing service value is a major component of the household s aggregate expenditure. For example, Zereyesus et al., (2017) indicate that in Northern Ghana following food expenditure, housing rent accounted was ranked the second highest category forming 16% of household s annual expenditure. Data shows that a significant proportion of households in a typical household survey adopted by the Word Bank are self-owned. Accurate measurement of housing services is therefore important to obtain true assessment of economic well-being. The standard approach of imputing the housing service values for such surveys is based on what it would cost to rent the house (Garner, Short and Kogan, 2006; Straszheim, 1974). This hedonic pricing is normally predicted using an Ordinary Least Squares (OLS) method. Such regression model has been widely applied in and out of sample predictions. OLS with minimum Mean Square Error (MSE) may offers the best fitted values for in-sample predictions. In OLS type of regressions, increasing the number of variables may increases the overall fit of the regression model (i.e. adjusted R 2 ), which in turn may increase the in-sample predictive power.. However, such complex model may not perform well in terms of predicting the out-of-sample dependent values.. Traditional models such as OLS are poorly suited for complex models (Song and Bickel, 2011), especially due to dimensionality. High-dimensional models comprise serial correlation, structural dependence among explanatory variables (spatial correlation), small sample size problem (small 3
4 sample size relative to large number of variables) (Belloni et al., 2012; Song and Bickel, 2011) or combination of all. In the world of high-dimensionality, OLS fails to select better predictors and leads to poor out-of-sample prediction. In our analysis, the accuracy of the housing service value for the majority of self-owned type of households is based on performance of the out-of-sample predictive power of the model. Machine learning approaches provide better out-of-sample predictive values considering higher dimensionality of data structures. As Vinod (1978), notes that experiments consistently supported a better prediction error by machine learning methods over OLS. Machine leaning approaches are widely used in computer science field and have recently been adopted in the statistics and economics fields. The current study aims to evaluate multiple machine learning approaches (i.e. Ridge, LASSO, Tree, Random forest, Bagging and Boosting) to predict housing service values using two period household level survey data from three sub-saharan countries (Uganda, Tanzania, and Malawi). The performance of these models are evaluated in reference to a standard OLS approach. Results show that in general the machine learning approaches were better in their predictive performance compared to an OLS approach. The different approaches vary from each other. Methodology According to Varian (2014), data analysis in statistics and econometrics can be classified into 1) prediction, 2) summarization, 3) estimation, and 4) hypothesis testing. The primary concern of machine learning is prediction. In machine learning, the focus is to find some function that provides a good prediction of Y as a function of X. The vector X is referred to as predictors or features. Machine learning methods are believed to perform better on modeling data with dimensionality problems such as serial correlation (Ng, 2013; Bai, and Ng, 2009; Andrews, 1991), spatial correlation (Caner and Fan, 2010; Lounici et al., 2009; Knight, 2008; Meinshausen and Bühlmann, 2006; Tibshirani, 1996), small sample size problem (Belloni et al., 2012; Huang, Horowitz and Wei, 2010; Meinshausen and Bühlmann, 2006) or combinational of all (Buehlmann, 2006). In general, these machine learning methods deal with high-dimensional data through shrinking some of the variables to zero and retaining the most important covariates. 4
5 The machine learning models generally partition the data into training and testing parts. The training data is used to build the model; while the testing data part is used to test the prediction power of the model. For convenience, the data in this study is partitions into two equal parts. A K-fold cross-validation is considered for model selection. In other applications, Akaike information criteria (AIC), Bayesian information criteria (BIC), and others can also be used for model selection. The advantage of cross validation over BIC, AIC and others is that it depends on fewer assumptions (Varian, 2014). Detailed procedure on how K-fold cross-validation works is contained in James et al. (2013). Following is a brief review of the standard OLS model and the machine learning approaches. The explanatory variables included in the OLS model and the features or factors in the machine learning models are selected based on their perceived relationship with the house rental price as supported by economic theory and housing economic literature (e.g. Garner, Short and Kogan, 2006; Graves et al., 1988; Straszheim, 1974). Ordinary Least Square model Assume that we have the following OLS model: Y = βx + ε, ε ~(0, σ 2 )... (1) Where Y is the outcome variable, X represents a matrix of explanatory variables, ε represents identically and independently distributed (IID) error term, and β refers to the matrix of parameters.. We run the linear regression model to find the optimal estimates that gives least squared errors. Minimum SE(β ) = E [(β β)(β β) ], where E is expected value. Bias-variance tradeoff Based on the OLS estimations of the variables coefficients, suppose that our fitted model is given by Y = β X. Relative to the actual values of Y, the Prediction error (PE) at particular point of X 0 is given by: PE(X 0 ) = E Y X=X0 {(Y Y ) 2 X = X 0 } 5
6 After some manipulation of the above prediction error equation, it can be decomposed into the following parts as: PE(X 0 ) = σ ε 2 + Bias 2 (Y (X 0 )) + Var(Y (X 0 )). This decomposition is called bias-variance tradeoff. As more terms are added to the model, the estimates are influenced by high variance. Hence, reducing some variables from the model may add little bias in the model but it may decrease the variance significantly, which in turn reduce the PE. This bias-variance decomposition is at the heart of machine learning models such as Ridge and LASSO (sometime referred as regression regularization methods) which attempt to introduce bias into the regression solution, but reduce the variance significantly as compared to the OLS solutions. The idea is that while OLS regressions provide unbiased coefficient estimates, the lower variance from such machine learning methods produces better Mean Squared Error results. Machine Learning Methods Ridge regression Ridge regression is similar to OLS in that it minimizes the MSE by applying a penalty parameter. The size of this parameter depends on the complexity of the model. Ridge model is specified as: Minimize i (Y i βx i ) λ j β j where λ represents penalizing parameter. At λ = 0, ridge equals to OLS. Least Absolute Shrinkage and Selection Operator Regression: The Least Absolute Shrinkage and Selection Operator (LASSO) and ridge are of the same family, where instead of penalizing square of coefficient in ridge, we penalize the summation of the absolute value of the coefficients in Lasso. The Lasso model is: Minimize i (Y i βx i ) 2 + λ j β j Unlike ridge, lasso reduces some coefficients to zero. Neither of them is superior to another. Ridge is chosen over Lasso when so many explanatory variables have small impact on the outcome 6
7 variable. Similarly, lasso is chosen over ridge when some explanatory variables have large effect on the outcome variable. Cross validation (Belloni, Chernozhukov and Hansen, 2012) is commonly used to select λ. Regression Tree: Unlike the previous machine learning methods, regression tree is a non-parametric approach that does not require specification of any particular functional form. It splits the data in to subtrees based on the variable that best explains the data. It keeps dividing the data in to subtrees until (1) there is single observation in each subtree; (2) all data in the subtree are identical, and (3) number of subtree can be determined by the practitioner. The model for the Regression tree is given by: Y = β n 1(Xε R n ) n The indicator 1(Xε R n ) is a dummy variable with a value of X =1 if the variable is with subtree n, and 0 otherwise. The β n is the mean value of the data in subtree n. Bagging Bagging (Breiman, 1996) is another type of regression tree applied to reduce potentially high variance using a bootstrapping method. As Breiman (1996) states, the bagging predictor is a method for generating multiple versions of a predictor to obtain an aggregated predictor. One of the drawbacks for bagging is that it may not perform well when the predictive power of some of the bagging regressions is much better than the others. Random Forests Random forest is similar procedure to bagging, however, unlike bagging; there is the possibility that less strong covariates could be chosen at random. Random forest performs better prediction when we have highly correlated explanatory variables. James et al. (2013) discuss that averaging various uncorrelated quantities offer much less variance compared to highly correlated quantities. 7
8 Inclusion of inappropriate variables and not sufficiently small selected variables in random forest results with poorer prediction. Boosting Boosting is another type of regression tree that follows slightly different mechanism from random forest. Unlike random forest that build trees independent of each other, boosting builds the trees sequentially. Each tree is built using the information from the previous created tree. Later, a tree is fitted to the residuals, instead of the outcome variable of the former tree and then the errors are updated by adding the new decision tree to the fitted function. This strategy avoids fitting large single tree in to the data that potentially reduces the possibility of overfitting the data. Data Data used in the study comes from the nationally representative Living Standards Measurement Study (LSMS) of World Bank household level surveys conducted in 2009/10, 2011/2012 and 2013/2014 in Uganda, Tanzania and Malawi, respectively. Information from surveyed households pertaining to rental values and other variables of interest describing the location, quality of housing, and other features were extracted from the survey data for use in the current study. The number of households that were rented out during the survey in each survey year by country form the basis of the sample sizes for the analyses. In Uganda, the sample sizes are 315 and 389 in the years 2009 and 2011, respectively. In Tanzania, the sample sizes are 699 and 956 in the years 2011 and 2013, respectively. In Malawi, the sample sizes are 1492 and 676 in the years 2011 and 2013, respectively (Table 1). The rental monetary values from the domestic currency in each of countries is converted to a US dollar to facilitate easier comparison and analyses of results. Table 1 presents descriptive statistics of key variables used in the analyses. In Uganda, households who live in urban area form 54 percent in years 2009 and 59 percent in Likewise, in Tanzania, over 70 percent of the households live in urban areas in both years, and in Malawi, households who live in urban area constitute 77 and 66 percent in years 2011 and 2013, respectively. The annual rent is almost constant over the years across all three countries. The rental rate per year in US dollars is around 300 in both years for Uganda and Malawi, where as in Tanzania, the rental rates are 168 in year 2009, and 260 in year The majority of households in all countries have roofs, floors, and external walls made up of mud. The average number of 8
9 rooms per a house ranged from 1.67 in Tanzania in year 2009 to 2.53 in Malawi in Percentage of households with access to electricity ranged from 37 percent to 54 percent. The type of water sources includes: private tap water, public tap water, bore-hole water, protected well water and unprotected water. There is no consistent pattern in the type of water source used in the three countries. Public tap water in Uganda (47%) and Malawi (27%) and protected well water in Tanzania (25%) are relatively dominant source of water used. The type of toilets available are covered private and shared toilets, VIP and uncovered latrine toilets, and flush toilets. The main type of toilet in both years is cover shared toilet (61%) in Uganda but it varies by year in Tanzania and Malawi. In the year 2013, VIP latrine type of toilet turned out to be the main type of toilet in Tanzania and Malawi. Results and Discussions Determinants of housing values Table 2-4 present the results from the OLS and various machine-learning approaches used in the study to estimate the determinants of housing service values in Uganda, Tanzania and Malawi. The significance of the explanatory variables in the OLS models are consistent in the two periods for the three countries. In Uganda, factors that are significantly associated with housing rental values are location in urban, number of rooms, availability of electricity, and private tap water. In Malawi, Urban location, number of rooms, electricity, and flash toilet are significantly associated with the housing rental values. In Tanzania, Urban location, number of rooms, external wall, availability of electricity, and protected well water are significantly associated with the house rental values. None of the other explanatory variables is significantly associated with the dependent variables in any of the three countries. Of the significantly associated variables, for example, Urban location is positively associate with rental values implying that rent is higher for houses located in urban than rural. On average, house rent in urban location is higher by $10-14 in Uganda, $5 in the year 2013 in Tanzania and $9 in the year 2011 in Malawi. Many households want to live in urban area because urban are associated with more employment opportunities, and better public services that lead to higher living standards (Sahn and Stifel, 2003). House rent increases with the increase of number of rooms. Increasing number of rooms by 1, increases on average monthly house rent by about $5 in Uganda, $7-13 in 9
10 Tanzania, and $11 in Malawi. The effect of electricity on house rent is also positive. Rental values of houses with access to electricity are higher by $19 per month in Uganda, $6-13 in Tanzania and $7 in the year 2013 in Malawi. These relationships suggest that electricity is essential for household s livelihood in that it can be used for household purposes (lighting, cooking, etc.), mechanizations (farming and non-farming) and communication and other purposes (Bernard, 2010; Khandker et al., 2009). Private tap water is also positively associated with the value of house rent in Uganda. On average, a house rents in dwellings with a private water tap are higher by $28 in Uganda. North and Griffin (1993) found that households value in-house piped water sources higher than any other house characteristics. Rental houses increases with access to clean water because more households want to stay closer to clean water (Nauges and Whittington, 2010; Quigley, 1982). In Tanzania, however, protected well type of water source is negatively associated with house rent. Results from the ridge machine learning regression models show that most of the estimated coefficients of the explanatory variables are smaller in magnitude compared to the OLS coefficients (see Table 2-4). This is consistent with the literature (James et al., 2013) in that ridge regression shrinks the estimated coefficients and reduces the variance resulting in increased predictive power of the covariates. Similarly, the LASSO regression approach shrinks the variable estimates and in some cases, the coefficients are reduced to zero. When the estimated variables coefficients are zero, then the variables may not play a role in predicting the value of the dependent variables. In Uganda. the variable estimates that were reduced to zero under LASSO regression are dwelling, floor, bore hole water, unprotected water sources, covered private and shared type of toilet, pit and uncovered latrine in 2009; and dwelling, roof, floor, public tap, bore hole and unprotected water sources, pit and uncovered latrine in In Tanzania variable estimates for roof, floor, public tap, protected and unprotected well water sources, VIP latrine type of toilet in 2011 and floor, public tap and unprotected well water sources, and VIP latrine type of toilet in 2013 are also reduced to zero under the LASSO regression. For Malawi, the variables with zero estimated coefficients under LASSO regression are dwelling, floor, public tap, bore hole, protected and unprotected well water sources and VIP latrine type of toilet in 2011 and dwelling, public tap, protected and unprotected well water sources, VIP and pit latrine type of toilet in 2013 in. 10
11 Unlike OLS, Ridge, and LASSO, the other machine learning approaches (i.e. the Regression trees, bagging, random forest and boosting) do not provide the estimated coefficients between the covariates and predicted variables. In regression tree, the prediction of the house rental values is done by using the mean/mode of the group at each node of the tree. For example, in Uganda, variables selected in the regression tree by rank are flush toilet, availability of electricity, protected water (well) and number of rooms in the year 2009; and electricity, number of rooms, protected well and private tap water sources in the year 2011 (Figure 1_u_a). The summary statistics indicate that only 4 percent of the households have access to flush toilet. The largest group, those without access to flush toilet are further split using electricity, protected water and number of rooms variables. Regression tree is highly criticized that changing the data slightly brings substantial change in tree construction (James et al., 2013). Likewise, variables used in the tree regression in Tanzania are number of rooms, electricity, pit latrine, private water tap, and external wall in the year 2009 and number of rooms, electricity, external wall, flush toilet and urban in the year In Malawi, variables such as flush toilet, electricity and number of rooms in 2011, and flush toilet, electricity and number of rooms in 2013 are used in the tree regression. As stated above, the bagging method is an extension of the regression tree. We construct numerous trees simply through resampling multiple times from the same data (i.e. bootstrapping) and then we average the entire prediction. In each bagging, 500 trees are created using twelve variables. The results from the bagging regression are sorted according to their significance in the prediction process. For example, in Uganda, the three top most variables used during the bagging regression by rank are flush toilet, protected water sources, and electricity; whereas the dwelling variables appears in the bottom most level (Figure 2_u_a). The arrangement of variables vary by year of analysis and country of analysis. Details of these results are presented in the appendix. Random forest is similar to bagging, but instead of averaging all the predictors, we take some of the predictors via random selection to perform the prediction. For example, out of the 12 bagging variables used, random forest picked 4 variables randomly. The top 4 variables selected in random forest in Uganda are private water tape sources, flush toilet, public water tap, and electricity in the 11
12 year 2009 (Figure 3_u_a). The rest of the results for the other models by year and by country is provided in the appendix. The results from the boosting regression, which is another extension to the regression tree are presented in table 5. The estimated from the boosting regression are arranged in order of their importance to the prediction process. For instance, In Uganda in the year 2009, the top three variables used in boosting by order of importance are electricity, number of rooms, and public water tape sources; the four of the variables that are not used for boosting are unprotected water sources, flush toilet, and roof. A number of variable are not used in the boosting regression of different countries and in different years as shown in the table. Out-of-sample prediction The main strength of machine learning approaches is their predictive capability. The Mean Squared Error is one of the methods used to compare between the machine learning approaches and the standard OLS regression. The results of such comparison of these models is presented in table 6. Our results indicate that the prediction power of machine learning is superior to OLS. But, it matters which type of machine learning we are comparing with OLS. For example, in Uganda in the year 2009, the prediction methods by rank are boosting, ridge, lasso, forest, OLS, bagging and tree; and bagging, forest, lasso, OLS, ridge, tree and boosting in In Uganda, OLS performed better in prediction than bagging and tree in 2009 and ridge, tree and boosting in However, OLS is less preferred than boosting, ridge, lasso and forest in the year 2009 and bagging, forest and lasso in the year In Tanzania, the prediction methods in order of preference judged by minimum MSE are forest, tree, boosting, OLS, lasso, ridge, and bagging in the year 2009 and forest, ridge, lasso, OLS, boosting, bagging and tree in the year This implies that forest, tree and boosting in the year 2009, and forest, ridge and lasso in the year 2013 performed better for out-of-sample prediction compared to OLS. On the other hand, OLS is superior to lasso, ridge and bagging in the year 2009, and boosting, bagging and tree in the year Importantly, the rank of the prediction methods from top to bottom in Malawi are forest, boosting, bagging, tree, OLS, ridge and lasso in the year 2011 and boosting, tree, bagging, forest, OLS, lasso, and ridge in the year Likewise, in 12
13 Malawi, in both years, OLS performed better in prediction than ridge and lasso but poorer prediction than boosting, bagging, tree and forest. Conclusion Generally, the results indicated that the machine learning methods have better prediction power than OLS. The use of cross-country data shows that the results vary by country as well as year of analysis. This is in agreement with James et al. (2013) conclusion that there is no method that dominates all other methods over any data set. Besides, the advantage of machine leaning models over OLS type regressions is significant when the number of predictor variables (features) is significant larger. 13
14 References: Andrews, D. W. (1991). Asymptotic optimality of generalized C L, cross-validation, and generalized cross-validation in regression with heteroskedastic errors. Journal of Econometrics, 47(2), Bajari, P., Nekipelov, D., Ryan, S. P., & Yang, M. (2015). Machine learning methods for demand estimation. The American Economic Review, 105(5), Bai, J., & Ng, S. (2009). Boosting diffusion indices. Journal of Applied Econometrics, 24(4), Belloni, A., Chen, D., Chernozhukov, V., & Hansen, C. (2012). Sparse models and methods for optimal instruments with an application to eminent domain. Econometrica, 80(6), Bernard, T. (2010). Impact analysis of rural electrification projects in sub-saharan Africa. The World Bank Research Observer, lkq008. Breiman, L Bagging Predictors Machine Learning 24 (2) : Buehlmann, P. (2006). Boosting for high-dimensional linear models. The Annals of Statistics, Caner, M., & Fan, Q. (2010). The adaptive lasso method for instrumental variable selection. Working Paper, North Carolina State University. Garner, T. I., Short, K. S., & Kogan, U. (2006). What Do We Know About the Value of Owner Occupied Housing Services? Rental Equivalence and Other Approaches. Graves, P., Murdoch, J. C., Thayer, M. A., & Waldman, D. (1988). The robustness of hedonic price estimation: urban air quality. Land Economics, 64(3), James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An introduction to statistical learning (Vol. 6). New York: springer. 14
15 Jimenez, E. (1982). The value of squatter dwellings in developing countries. Economic Development and Cultural Change, 30(4), Khandker, S. R., Barnes, D. F., Samad, H. A., & Minh, N. H. (2009). Welfare impacts of rural electrification: evidence from Vietnam. World Bank Policy Research Working Paper, (5057). Hastie, T., Tibshirani, R., & Friedman, J. (2009). Unsupervised learning. In The elements of statistical learning (pp ). Springer New York. Huang, J., Horowitz, J. L., & Wei, F. (2010). Variable selection in nonparametric additive models. Annals of statistics, 38(4), Knight, K. (2008). Shrinkage estimation for nearly singular designs. Econometric Theory, 24(02), Lounici, K., Pontil, M., Tsybakov, A. B., & Van De Geer, S. (2009). Taking advantage of sparsity in multi-task learning. arxiv preprint arxiv: Meinshausen, N., & Bühlmann, P. (2006). High-dimensional graphs and variable selection with the lasso. The annals of statistics, Nauges, C., & Whittington, D. (2010). Estimation of water demand in developing countries: An overview. The World Bank Research Observer, 25(2), Ng, S. (2013). Variable selection in predictive regressions. Handbook of Economic Forecasting, 2(Part B), North, J. H., & Griffin, C. C. (1993). Water source as a housing characteristic: Hedonic property valuation and willingness to pay for water. Water Resources Research, 29(7), Quigley, J. M. (1982). Nonlinear budget constraints and consumer demand: An application to public programs for residential housing. Journal of Urban Economics, 12(2), Sahn, D. E., & Stifel, D. C. (2003). Urban rural inequality in living standards in Africa. Journal of African Economies, 12(4), Song, S., & Bickel, P. J. (2011). Large vector auto regressions. arxiv preprint arxiv:
16 Straszheim, M. (1974). Hedonic estimation of housing market prices: A further comment. The Review of Economics and Statistics, Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society. Series B (Methodological), Zereyesus, Y. A., Embaye, W. T., Tsiboe, F., & Amanor-Boadu, V. (2017). Implications of Non-Farm Work to Vulnerability to Food Poverty-Recent Evidence from Northern Ghana. World Development, 91, Varian, H. R. (2014). Big data: New tricks for econometrics. The Journal of Economic Perspectives, 28(2), Vinod, H. D. (1978). A survey of ridge regression and related techniques for improvements over ordinary least squares. The Review of Economics and Statistics,
17 Tables and figures: Table 1: Summary statistics of rental house characteristics Variable Definitions Mean values Ugandan Tanzania Malawi Urban 1 if a house is located in Urban, 0 otherwise 0.54 (0.50) 0.59 (0.49) 0.94 (0.45) 0.75 (0.43) 0.77 (0.42) 0.66 (0.47) Annual rent Amount of rent paid per year in dollar 290 (37.19) 290 (37.19) 168 (35.24) 260 (37.06) 300 (50.15) 300 (44.51) Dwelling 1 if the room is located within a house, 0 otherwise 0.69 (0.46) 0.70 (0.46) - (0.49) 0.45 (0.49) 0.87 (0.34) 0.92 (0.28) Roof 1 if roof type if mud, 0 otherwise 0.92 (0.28) 0.92 (0.26) 0.96 (0.23) 0.93 (0.25) 0.82 (0.39) 0.86 (0.35) floor 1 if floor is mud, 0 otherwise 0.70 (0.46) 0.70 (0.46) 0.97 (0.41) 0.77 (0.41) 0.74 (0.44) 0.79 (0.41) External wall 1 if external wall is in mud, 0 otherwise 0.72 (0.45) 0.74 (0.44) 0.67 (0.47) 0.72 (0.45) 0.05 (0.22) 0.05 (0.21) Number of rooms Number of rooms rented 1.80 (1.11) 1.77 (1.16) 1.67 (0.97) 1.68 (0.91) 2.53 (1.09) 2.11 (0.97) Electricity 1 if the house has electricity, 0 otherwise 0.37 (0.48) 0.41 (0.49) 0.54 (0.50) 0.54 (0.50) 0.37 (0.48) 0.43 (0.50) Water sources Private tap water 1 if private tap water source, 0 otherwise 0.08 (0.27) 0.07 (0.25) 0.35 (0.26) 0.12 (0.32) 0.13 (0.34) 0.14 (0.35) 17
18 Variable Definitions Mean values Ugandan Tanzania Malawi Public tap water 1 if public tap water source, 0 otherwise 0.49 (0.50) 0.47 (0.49) 0.39 (0.38) 0.12 (0.33) 0.27 (0.44) 0.27 (0.44) Bore hole water 1 if bore hole water source, 0 otherwise 0.19 (0.39) 0.19 (0.39) (0.25) 0.33 (0.47) 0.31 (0.46) Protected well water 1 if protected well water source, 0 otherwise 0.14 (0.34) 0.13 (0.34) 0.11 (0.43) 0.25 (0.43) 0.04 (0.22) 0.04 (0.28) Unprotected water 1 if unprotected well water, 0 otherwise 0.10 (0.34) 0.10 (0.25) 0.15 (0.24) 0.13 (0.34) 0.21 (0.41) 0.20 (0.40) Covered private toilet 1 if covered private toilet, 0 otherwise 0.16 (0.37) 0.17 (0.37) (0.19) 0.14 (0.35) 0.16 (0.37) Cover shared toilet 1 if covered shared toilet, 0 otherwise 0.61 (0.49) 0.62 (0.48) (0.43) 0.09 (0.29) 0.05 (0.22) VIP latrine toilet 1 if VIP latrine toilet, 0 otherwise 0.07 (0.25) 0.08 (0.26) 0.17 (0.41) 0.05 (0.22) 0.67 (0.47) 0.64 (0.48) Uncovered latrine 1 if uncovered latrine, 0 otherwise 0.10 (0.29) (0.40) - (0.27) Pit latrine 1 if pit latrine, 0 otherwise (0.45) Flush toilet 1 if flush toilet, 0 otherwise 0.04 (0.19) 0.13 (0.33) Observation
19 Table 2: Determinants of housing rental values based on OLS, Lasso and Ridge regressions models in Uganda Variables OLS Ridge Lasso OLS Ridge Lasso Constant (42.03) (9.12) Urban 13.92* * (4.02) (2.62) Dwelling (4.50) (3.06) Roof (8.56) (5.68) floor (5.47) External wall 9.93 (5.36) Number of rooms 5.75* (1.79) Electricity 18.80** (4.05) Private tap water (30.43) Public tap water 4.59 (29.74) Bore hole water (29.96) Protected well water (29.98) Unprotected water (30.29) Covered private toilet (27.94) Cover shared toilet 2.60 (25.99) VIP latrine toilet (32.08) Flush toilet (30.43) (3.74) (3.65) *** (1.04) * (2.91) ** (6.79) (3.96) (4.21) (7.35) (7.96) (12.01) (7.99) (12.16) (25.68) Pit latrine Uncovered latrine (26.04) (8.73) R
20 Table 3: Determinants of housing rental values based on OLS, Lasso and Ridge regressions models in Tanzania Variables OLS Ridge Lasso OLS Ridge Lasso Constant ** (16.34) (6.12) Urban 5.55** * (5.27) (4.42) Dwelling Roof 3.94 (20.91) floor (22.45) External wall 7.92** (3.15) Number of rooms 6.93*** (1.39) (7.45) (5.09) *** (3.68) *** (1.76) Electricity 5.90* *** (3.05) (3.83) Private tap water (4.43) (3.20) Public tap water (3.97) (6.59) Bore hole water Protected well water * (4.56) (8.31) Unprotected water (4.03) Covered private toilet Cover shared toilet VIP latrine toilet 3.64 (15.79) Flush toilet (15.94) Pit latrine (15.40) (7.54) (6.64) (6.39) Uncovered latrine R
21 Table 4: Determinants of housing rental values based on OLS, Lasso and Ridge regressions models in Malawi Variables OLS Ridge Lasso OLS Ridge Lasso Constant (36.76) (9.53) Urban 8.96** (4.43) (4.93) Dwelling (10.94) (11.67) Roof (8.54) (9.13) floor (5.58) (5.30) External wall 9.67 (9.05) (7.76) Number of rooms 10.54*** (1.98) *** (1.87) Electricity 6.80** * (5.04) (4.68) Private tap water (21.98) Public tap water (20.32) (18.97) Bore hole water (12.27) (19.21) Protected well water (13.44) (26.37) Unprotected water (21.23) (18.93) Covered private toilet Cover shared toilet VIP latrine toilet (15.09) (16.61) Flush toilet (35.60) Pit latrine 6.81 (35.32) *** (20.61) (12.39) Uncovered latrine R
22 Figure 1: Tree Regression in Uganda in year 2009 Flush toilet No Yes electricity No Yes Protected Water No Yes No. rooms < No Yes
23 Bagging regression results Figure 2_u_a figure 2_u_b Figure 2_t_a Figure 2_t_b 23
24 Figure 2_m_a Figure 2_m_b 24
25 Forest regression results Figure 2_u_a Figure 2_u_b 25
26 Figure 2_t_a Figure 2_t_b Figure 2_m_a Figure 2_m_b 26
27 Table: Table 5: Modeling factors associated to housing rental values using boosting regression Uganda Tanzania Malawi Variable Rel. info Variable Rel. info Variable Rel. info Variable Rel. info Variable Rel. info Variable Rel. info Electricity Number Number of rooms Number of Flush toilet Flush toilet 4.97 Of rooms rooms Number Electricity Flush toilet Electricity Number of Number of 3.94 Of rooms rooms rooms Public tap water Protected well Pit latrine External wall Electricity 4.65 Electricity 6.76 Protected well 9.57 Urban 7.14 Private tap water Flush toilet 2.76 Urban 2.03 Private tap 2.60 water Urban 6.83 Public tap water 4.10 Electricity 5.96 Pit latrine 2.71 Pit latrine 1.32 External 0.60 wall Private tap water 5.28 Dwelling 3.69 External wall 2.81 Urban 2.53 Private tap 0.93 Pit latrine 0.35 water Dwelling 3.99 Shared toilet 2.60 VIP toilet 2.29 Private tap 1.68 roof 0.41 roof 0.33 water Shared toilet 3.67 External wall 1.74 Public tap water 0.23 Public tap 0.42 External wall 0.36 floor 0.15 water External wall 1.49 floor 1.29 Urban 0.22 Protected 0.25 floor 0.16 Bore hole 0.10 well floor 1.04 Private tap water 0.71 Unprotected water 0.02 VIP toilet 0.24 VIP toilet 0.04 Urban 0.07 Bore hole 0.73 Private toilet 0.53 Protected well 0.02 roof 0.07 Bore hole 0.01 Dwelling 0.01 VIP toilet 0.55 Flush toilet 0.39 roof 0.00 floor 0.02 Dwelling 0.00 Public tap 0.01 water Private toilet 0.31 Unprotected water 0.06 floor 0.00 Unprotected 0.02 Protected well 0.00 Protected 0.00 water well Uncovered toilet 0.10 Bore hole Public tap 0.00 VIP toilet 0.00 water Unprotected water 0.00 Uncovered toilet Unprotected 0.00 Unprotected 0.00 water water Flush toilet 0.00 VIP toilet No toilet 0.00 No toilet 0.00 roof 0.00 roof
28 Table 6: Prediction mean square error of estimating housing values under various models Ugandan Tanzania Malawi OLS Ridge Lasso Tree Bagging Forest Boosting
Estimating Demand Elasticities of Meat Demand in Slovakia
Estimating Demand Elasticities of Meat Demand in Slovakia Daniela Hupkova (1) - Peter Bielik (2) (1) (2) Slovak University of Agriculture in Nitra, Faculty of Economics and Management, Department of Economics,
More informationEvaluation of random forest regression for prediction of breeding value from genomewide SNPs
c Indian Academy of Sciences RESEARCH ARTICLE Evaluation of random forest regression for prediction of breeding value from genomewide SNPs RUPAM KUMAR SARKAR 1,A.R.RAO 1, PRABINA KUMAR MEHER 1, T. NEPOLEAN
More informationToday. Last time. Lecture 5: Discrimination (cont) Jane Fridlyand. Oct 13, 2005
Biological question Experimental design Microarray experiment Failed Lecture : Discrimination (cont) Quality Measurement Image analysis Preprocessing Jane Fridlyand Pass Normalization Sample/Condition
More informationExamination of Cross Validation techniques and the biases they reduce.
Examination of Cross Validation techniques and the biases they reduce. Dr. Jon Starkweather, Research and Statistical Support consultant. The current article continues from last month s brief examples
More informationRetail Sales Forecasting at Walmart. Brian Seaman WalmartLabs
Retail Sales Forecasting at Walmart Brian Seaman WalmartLabs 1 The biggest challenge as a forecasting practitioner The boss says: I need a forecast of A forecaster should respond: Why? 2 Today s Focus
More informationTechnical note on the determinants of water and sanitation in Yemen
Technical note on the determinants of water and sanitation in Yemen 1. Introduction Giacomo Sbrana 1 Department of Economic and Social Affairs (DESA) United Nations, New York 02 June 2009 This technical
More informationTop-down Forecasting Using a CRM Database Gino Rooney Tom Bauer
Top-down Forecasting Using a CRM Database Gino Rooney Tom Bauer Abstract More often than not sales forecasting in modern companies is poorly implemented despite the wealth of data that is readily available
More informationReal-Time Predictive Modeling of Key Quality Characteristics Using Regularized Regression: SAS Procedures GLMSELECT and LASSO
Paper 8240-2016 Real-Time Predictive Modeling of Key Quality Characteristics Using Regularized Regression: SAS Procedures GLMSELECT and LASSO Jon M. Lindenauer, Weyerhaeuser Company ABSTRACT This paper
More informationSeasonality in local food markets. Africa IN AFRICA. J. Kaminski, L. Christiaensen, C. Gilbert, and C. Udry
Seasonality in local food markets and household consumption in Africa J. Kaminski, L. Christiaensen, C. Gilbert, and C. Udry IMF-OCP-NYU Conference on food price volatility in LDCs Rabat, Maroocoo February,
More informationBOOSTED TREES FOR ECOLOGICAL MODELING AND PREDICTION
Ecology, 88(1), 2007, pp. 243 251 Ó 2007 by the Ecological Society of America BOOSTED TREES FOR ECOLOGICAL MODELING AND PREDICTION GLENN DE ATH 1 Australian Institute of Marine Science, PMB 3, Townsville
More informationWinsor Approach in Regression Analysis. with Outlier
Applied Mathematical Sciences, Vol. 11, 2017, no. 41, 2031-2046 HIKARI Ltd, www.m-hikari.com https://doi.org/10.12988/ams.2017.76214 Winsor Approach in Regression Analysis with Outlier Murih Pusparum Qasa
More informationSPM 8.2. Salford Predictive Modeler
SPM 8.2 Salford Predictive Modeler SPM 8.2 The SPM Salford Predictive Modeler software suite is a highly accurate and ultra-fast platform for developing predictive, descriptive, and analytical models from
More informationApplying Regression Techniques For Predictive Analytics Paviya George Chemparathy
Applying Regression Techniques For Predictive Analytics Paviya George Chemparathy AGENDA 1. Introduction 2. Use Cases 3. Popular Algorithms 4. Typical Approach 5. Case Study 2016 SAPIENT GLOBAL MARKETS
More informationHouseholds Choice of Drinking Water Sources in Malawi
Households Choice of Drinking Water Sources in Malawi Presented by Stevier Kaiyatsa on behalf of Lawrence Mapemba and Gelson Tembo 12 th Meeting of the International Water Resource Economics Consortium
More informationBig Data. Methodological issues in using Big Data for Official Statistics
Giulio Barcaroli Istat (barcarol@istat.it) Big Data Effective Processing and Analysis of Very Large and Unstructured data for Official Statistics. Methodological issues in using Big Data for Official Statistics
More informationThe methods to estimate the monetary value of the environment
The methods to estimate the monetary value of the environment Dr. Alberto Longo Department of Economics and International Development University of Bath A.Longo@bath.ac.uk 1 Overview of the presentation
More informationESTIMATING GENDER DIFFERENCES IN AGRICULTURAL PRODUCTIVITY: BIASES DUE TO OMISSION OF GENDER-INFLUENCED VARIABLES AND ENDOGENEITY OF REGRESSORS
ESTIMATING GENDER DIFFERENCES IN AGRICULTURAL PRODUCTIVITY: BIASES DUE TO OMISSION OF GENDER-INFLUENCED VARIABLES AND ENDOGENEITY OF REGRESSORS by Nina Lilja, Thomas F. Randolph and Abrahmane Diallo* Selected
More informationCOORDINATING DEMAND FORECASTING AND OPERATIONAL DECISION-MAKING WITH ASYMMETRIC COSTS: THE TREND CASE
COORDINATING DEMAND FORECASTING AND OPERATIONAL DECISION-MAKING WITH ASYMMETRIC COSTS: THE TREND CASE ABSTRACT Robert M. Saltzman, San Francisco State University This article presents two methods for coordinating
More informationWhitepaper Hybrid Econometric Model
Whitepaper Hybrid Econometric Model The hybrid model could be used to evaluate economic impact. Examples include infrastructure projects, special events and tourism. By Sanjeev Naguleswaran and John Colombo
More informationEffects of Livelihood Assets on Poverty Status of Farming Households in Southwestern, Nigeria
Effects of Livelihood Assets on Poverty Status of Farming Households in Southwestern, Nigeria LAWAL, J.O, 1 OMONONA B.T. 2 AND OYINLEYE, O.D 2 1 Economics Section, Cocoa Research Institute of Nigeria,
More informationThe Instability of Unskilled Earnings
XII. LABOR ECONOMICS/LABOR MARKETS AND HUMAN RESOURCES REFEREED PAPERS The Instability of Unskilled Earnings Michael Mamo Westminster College Wei-Chiao Huang Western Michigan University Abstract The year-to-year
More informationPredicting user rating on Amazon Video Game Dataset
Predicting user rating on Amazon Video Game Dataset CSE190A Assignment2 Hongyu Li UC San Diego A900960 holi@ucsd.edu Wei He UC San Diego A12095047 whe@ucsd.edu ABSTRACT Nowadays, accurate recommendation
More informationLeveraging Smart Meter Data & Expanding Services BY ELLEN FRANCONI, PH.D., BEMP, MEMBER ASHRAE; DAVID JUMP, PH.D., P.E.
ASHRAE www.ashrae.org. Used with permission from ASHRAE Journal. This article may not be copied nor distributed in either paper or digital form without ASHRAE s permission. For more information about ASHRAE,
More informationWage Mobility within and between Jobs
Wage Mobility within and between Jobs Peter Gottschalk 1 April 2001 Abstract This paper presents evidence on the extent of wage mobility both while working for the same firm and when moving to a new firm.
More informationTree Depth in a Forest
Tree Depth in a Forest Mark Segal Center for Bioinformatics & Molecular Biostatistics Division of Bioinformatics Department of Epidemiology and Biostatistics UCSF NUS / IMS Workshop on Classification and
More informationGoogle Data in Bridge Equation Models for German GDP. Thomas B. Götz* Deutsche Bundesbank, Frankfurt am Main, Germany -
Google Data in Bridge Equation Models for German GDP Thomas B. Götz* Deutsche Bundesbank, Frankfurt am Main, Germany - thomas.goetz@bundesbank.de Thomas A. Knetsch Deutsche Bundesbank, Frankfurt am Main,
More informationBox Office Buzz: Does Social Media Data Steal the Show from Model Uncertainty When Forecasting for Hollywood?
Box Office Buzz: Does Social Media Data Steal the Show from Model Uncertainty When Forecasting for Hollywood? Steven Lehrer and Tian Xie Queen s University, NYU Shanghai, and NBER, lehrers@queensu.ca Xiamen
More informationRegional Development and Inequality of Income Distribution
Regional Development and Inequality of Income Distribution Nasfi Fkili Wahiba Doctor of Economics, Research Unit "Enterprise Economy Environment" Higher Institute of Management University of Gabes, Tunisia
More informationMeasuring Household Resilience to Food Insecurity
ERD Accra, 21-23 May 2009 Measuring Household Resilience to Food Insecurity An Application to Palestinian Households Luca Alinovi, ESAF Erdgin Mane, ESAF Donato Romano, University of Florence Research
More informationEvaluation of the Residential Inclining Block Rate F2009-F2012. Revision 2. June Prepared by: BC Hydro Power Smart Evaluation
Evaluation of the Residential Inclining Block Rate F2009-F2012 Revision 2 June 2014 Prepared by: BC Hydro Power Smart Evaluation Table of Contents Executive Summary... ii 1. Introduction... 1 1.1. Evaluation
More informationA SIMULATION STUDY OF THE ROBUSTNESS OF THE LEAST MEDIAN OF SQUARES ESTIMATOR OF SLOPE IN A REGRESSION THROUGH THE ORIGIN MODEL
A SIMULATION STUDY OF THE ROBUSTNESS OF THE LEAST MEDIAN OF SQUARES ESTIMATOR OF SLOPE IN A REGRESSION THROUGH THE ORIGIN MODEL by THILANKA DILRUWANI PARANAGAMA B.Sc., University of Colombo, Sri Lanka,
More informationWeek 10: Heteroskedasticity
Week 10: Heteroskedasticity Marcelo Coca Perraillon University of Colorado Anschutz Medical Campus Health Services Research Methods I HSMP 7607 2017 c 2017 PERRAILLON ARR 1 Outline The problem of (conditional)
More informationARIMA LAB ECONOMIC TIME SERIES MODELING FORECAST Swedish Private Consumption version 1.1
Bo Sjo 2011-11-10 (Updated) ARIMA LAB ECONOMIC TIME SERIES MODELING FORECAST Swedish Private Consumption version 1.1 Send in a written report to bosjo@liu.se before Wednesday November 25, 2012. 1 1. Introduction
More informationDATA ANALYTICS WITH R, EXCEL & TABLEAU
Learn. Do. Earn. DATA ANALYTICS WITH R, EXCEL & TABLEAU COURSE DETAILS centers@acadgild.com www.acadgild.com 90360 10796 Brief About this Course Data is the foundation for technology-driven digital age.
More informationKeyword Performance Prediction in Paid Search Advertising
Keyword Performance Prediction in Paid Search Advertising Sakthi Ramanathan 1, Lenord Melvix 2 and Shanmathi Rajesh 3 Abstract In this project, we explore search engine advertiser keyword bidding data
More informationNon-Farm Enterprises and Poverty Reduction amongst Households in Rural Nigeria: A Propensity Score Matching Approach
IOSR Journal Of Humanities And Social Science (IOSR-JHSS) Volume 19, Issue 4 Ver. VI (Apr. 2014), PP 57-61 e-issn: 2279-0837, p-issn: 2279-0845. www.iosrjournals.org Non-Farm Enterprises and Poverty Reduction
More informationSurvey Statistician to provide assistance for the Randomized rural household survey Scope of Work (SOW)
AgResults Kenya On-Farm Storage Pilot Survey Statistician to provide assistance for the Randomized rural household survey Scope of Work (SOW) 1. Consultant Name: TBD 2. Period of Performance: TBD 3. Level
More informationEnsemble Modeling. Toronto Data Mining Forum November 2017 Helen Ngo
Ensemble Modeling Toronto Data Mining Forum November 2017 Helen Ngo Agenda Introductions Why Ensemble Models? Simple & Complex ensembles Thoughts: Post-real-life Experimentation Downsides of Ensembles
More informationMARKET VALUE AND PATENTS A Bayesian Approach. Mark HIRSCHEY * 1. Introduction
Economics Letters 27 (1988) 83-87 North-Holland 83 MARKET VALUE AND PATENTS A Bayesian Approach Robert A. CONNOLLY University of Cal~fornm, Irvine, CA 92717, USA Mark HIRSCHEY * Unmwity of Kansas, Lawrence,
More informationAre the youth exiting agriculture en masse? Eugenie Maïga, Luc Christiaensen, and Amparo Palacios-Lopez. September, 2015 version.
Are the youth exiting agriculture en masse? Eugenie Maïga, Luc Christiaensen, and Amparo Palacios-Lopez September, 2015 version Abstract This paper investigates the extent of youth engagement in agriculture
More informationThe relationship between innovation and economic growth in emerging economies
Mladen Vuckovic The relationship between innovation and economic growth in emerging economies 130 - Organizational Response To Globally Driven Institutional Changes Abstract This paper will investigate
More informationHow to estimate the model of sustainable profit and corporate social responsibility
Peer-reviewed and Open access journal ISSN: 1804-5006 www.academicpublishingplatforms.com The primary version of the journal is the on-line version BEH - Business and Economic Horizons Volume 9 Issue 4
More informationJournal of Asian Scientific Research
Journal of Asian Scientific Research journal homepage: http://aessweb.com/journal-detail.php?id=5003 A METAFRONTIER PRODUCTION FUNCTION FOR ESTIMATION OF TECHNICAL EFFICIENCIES OF WHEAT FARMERS UNDER DIFFERENT
More informationIMPLICATIONS OF SEASONAL PRICE AND PRODUCTIVITY CHANGES AT THE HOUSEHOLD LEVEL IN UGANDA - A HETEROGENEOUS AGENT APPROACH
IMPLICATIONS OF SEASONAL PRICE AND PRODUCTIVITY CHANGES AT THE HOUSEHOLD LEVEL IN UGANDA - A HETEROGENEOUS AGENT APPROACH Mark Musumba Agriculture and Food Security Center Earth Institute at Columbia University
More informationMeta-regression Analysis of Agricultural Productivity Studies
Meta-regression Analysis of Agricultural Productivity Studies Giannis Karagiannis Selected Paper prepared for presentation at the International Agricultural Trade Research Consortium s (IATRC s) 2013 Symposium:
More informationData Mining. Chapter 7: Score Functions for Data Mining Algorithms. Fall Ming Li
Data Mining Chapter 7: Score Functions for Data Mining Algorithms Fall 2011 Ming Li Department of Computer Science and Technology Nanjing University The merit of score function Score function indicates
More informationTreatment of Influential Values in the Annual Survey of Public Employment and Payroll
Treatment of Influential s in the Annual Survey of Public Employment and Payroll Joseph Barth, John Tillinghast, and Mary H. Mulry 1 U.S. Census Bureau joseph.j.barth@census.gov Abstract Like most surveys,
More informationA World Model of the Demand for
ForestSci., Vol. 33, No. 1, 1987, pp. 185-196 Copyright 1987, by the Society of American Foresters A World Model of the Demand for Paper and Paperboard ANDERS BAUDIN LARS LUNDBERG A STgACT. The objective
More informationHarbingers of Failure: Online Appendix
Harbingers of Failure: Online Appendix Eric Anderson Northwestern University Kellogg School of Management Song Lin MIT Sloan School of Management Duncan Simester MIT Sloan School of Management Catherine
More informationUse of a delayed sampling design in business surveys coordinated by permanent random numbers
Use of a delayed sampling design in business surveys coordinated by permanent random numbers Fredrik Jonsson, Eva Elvers, Jörgen Brewitz 1 Abstract 2 Sample coordination is used to increase, or decrease,
More informationKuhn-Tucker Estimation of Recreation Demand A Study of Temporal Stability
Kuhn-Tucker Estimation of Recreation Demand A Study of Temporal Stability Subhra Bhattacharjee, Catherine L. Kling and Joseph A. Herriges Iowa State University Contact: subhra@iastate.edu Selected Paper
More informationEstimating the drivers of regional rural land use change in New Zealand
Estimating the drivers of regional rural land use change in New Zealand Suzi Kerr, Jo Hendy, Kelly Lock and Yun Liang Motu Economic and Public Policy Research, PO Box 24390, Manners Street, 6142, Wellington
More informationDifferences Between High-, Medium-, and Low-Profit Cow-Calf Producers: An Analysis of Kansas Farm Management Association Cow-Calf Enterprise
Differences Between High-, Medium-, and Low-Profit Cow-Calf Producers: An Analysis of 2010-2014 Kansas Farm Management Association Cow-Calf Enterprise Dustin L. Pendell (dpendell@ksu.edu), Youngjune Kim
More informationEffect of Regulatory Quality on Wood Pulp Imports by India
Effect of Regulatory Quality on Wood Pulp Imports by India Joy Das Graduate Research Assistant Department of Agricultural Economics and Agribusiness Louisiana State University Email: jdas2@lsu.edu Phone(M):
More informationDifferences Between High-, Medium-, and Low-Profit Cow-Calf Producers: An Analysis of Kansas Farm Management Association Cow-Calf Enterprise
Differences Between High-, Medium-, and Low-Profit Cow-Calf Producers: An Analysis of 2012-2016 Kansas Farm Management Association Cow-Calf Enterprise Dustin L. Pendell (dpendell@ksu.edu) and Kevin L.
More informationAgriculture and Climate Change Revisited
Agriculture and Climate Change Revisited Anthony Fisher 1 Michael Hanemann 1 Michael Roberts 2 Wolfram Schlenker 3 1 University California at Berkeley 2 North Carolina State University 3 Columbia University
More informationWRITTEN PRELIMINARY Ph.D. EXAMINATION. Department of Applied Economics. University of Minnesota. June 16, 2014 MANAGERIAL, FINANCIAL, MARKETING
WRITTEN PRELIMINARY Ph.D. EXAMINATION Department of Applied Economics University of Minnesota June 16, 2014 MANAGERIAL, FINANCIAL, MARKETING AND PRODUCTION ECONOMICS FIELD Instructions: Write your code
More informationMultidimensions of Urban Poverty: Evidence from India
WP-2007-008 Multidimensions of Urban Poverty: Evidence from India S Chandrasekhar and Abhiroop Mukhopadhyay Indira Gandhi Institute of Development Research, Mumbai June 2007 Multidimensions of Urban Poverty:
More informationE-Commerce Sales Prediction Using Listing Keywords
E-Commerce Sales Prediction Using Listing Keywords Stephanie Chen (asksteph@stanford.edu) 1 Introduction Small online retailers usually set themselves apart from brick and mortar stores, traditional brand
More informationDeterminants of changes in youth and women agricultural labor participation in. selected African countries. Eugenie W. H. Maiga
Determinants of changes in youth and women agricultural labor participation in selected African countries Eugenie W. H. Maiga Assistant Professor, Université de Koudougou, Burkina Faso eugeniemaiga@gmail.com
More informationNON-MARKET VALUATION OF RURAL TOURISM IN YOGYAKARTA, INDONESIA
NON-MARKET VALUATION OF RURAL TOURISM IN YOGYAKARTA, INDONESIA Endah Saptutyningsih, Nugroho Joko Prasetyo, Esty Setyaningrum Faculty of Economics Universitas Muhammadiyah Yogyakarta Jalan Lingkar Selatan,
More informationPopulation Growth and Land Scarcity in Rwanda: The other side of the Coin
Population Growth and Land Scarcity in Rwanda: The other side of the Coin Alfred R. BIZOZA (PhD) Agricultural Economist,University of Rwanda 2014 Conference on Land Policy in Africa, Addis Ababa, Ethiopia
More informationALL POSSIBLE MODEL SELECTION IN PROC MIXED A SAS MACRO APPLICATION
Libraries Annual Conference on Applied Statistics in Agriculture 2006-18th Annual Conference Proceedings ALL POSSIBLE MODEL SELECTION IN PROC MIXED A SAS MACRO APPLICATION George C J Fernandez Follow this
More informationSPATIAL ANALYSIS OF DAIRY YIELDS RESPONSE TO INTENSIVE FARMING IN NEW ZEALAND
Yang, W. & Sharp, B. 2016. Spatial analysis of dairy yields response to intensive farming in New Zealand. In: Integrated nutrient and water management for sustainable farming. (Eds L.D. Currie and R. Singh).
More informationAsk the Expert Model Selection Techniques in SAS Enterprise Guide and SAS Enterprise Miner
Ask the Expert Model Selection Techniques in SAS Enterprise Guide and SAS Enterprise Miner SAS Ask the Expert Model Selection Techniques in SAS Enterprise Guide and SAS Enterprise Miner Melodie Rush Principal
More information2 Maria Carolina Monard and Gustavo E. A. P. A. Batista
Graphical Methods for Classifier Performance Evaluation Maria Carolina Monard and Gustavo E. A. P. A. Batista University of São Paulo USP Institute of Mathematics and Computer Science ICMC Department of
More informationLet the data speak: Machine learning methods for data editing and imputation
Working Paper 31 UNITED NATIONS ECONOMIC COMMISSION FOR EUROPE CONFERENCE OF EUROPEAN STATISTICIANS Work Session on Statistical Data Editing (Budapest, Hungary, 14-16 September 2015) Topic (v): Emerging
More informationModelling and Forecasting the Balance of Trade in Ethiopia
American Journal of Theoretical and Applied Statistics 2015; 4(1-1): 19-23 Published online March 18, 2015 (http://www.sciencepublishinggroup.com/j/ajtas) doi: 10.11648/j.ajtas.s.2015040101.14 ISSN: 2326-8999
More informationDecision Tree Learning. Richard McAllister. Outline. Overview. Tree Construction. Case Study: Determinants of House Price. February 4, / 31
1 / 31 Decision Decision February 4, 2008 2 / 31 Decision 1 2 3 3 / 31 Decision Decision Widely Used Used for approximating discrete-valued functions Robust to noisy data Capable of learning disjunctive
More informationMachine Learning for Causal Inference: An Application to Air Quality Impacts on House Prices
Machine Learning for Causal Inference: An Application to Air Quality Impacts on House Prices Jenny Ho University of Washington October 27, 2016 Latest Version Abstract The increase in availability of big
More informationLocal Impacts of Resource Shocks Three quantitative case studies from Indonesia
Local Impacts of Resource Shocks Three quantitative case studies from Indonesia Ryan B. Edwards Contributed presentation at the 60th AARES Annual Conference, Canberra, ACT, 2-5 February 2016 Copyright
More informationMegatrends Driving Agricultural Transformation in Africa
Megatrends Driving Agricultural Transformation in Africa Focus on Malawi T. S. Jayne, Kwame Yeboah, and Milu Muyanga Presentation at USAID/Malawi, Lilongwe, Malawi 24 July, 2017 Five inter-related trends
More informationStartup Machine Learning: Bootstrapping a fraud detection system. Michael Manapat
Startup Machine Learning: Bootstrapping a fraud detection system Michael Manapat Stripe @mlmanapat About me: Engineering Manager of the Machine Learning Products Team at Stripe About Stripe: Payments infrastructure
More informationPredicting Pavement Condition Index Using International Roughness Index in a Dense Urban Area
Journal of Civil Engineering Research 2015, 5(1): 10-17 DOI: 10.5923/j.jce.20150501.02 Predicting Pavement Condition Index Using International Roughness Index in a Dense Urban Area Stephen A. Arhin 1,*,
More informationModeling of competition in revenue management Petr Fiala 1
Modeling of competition in revenue management Petr Fiala 1 Abstract. Revenue management (RM) is the art and science of predicting consumer behavior and optimizing price and product availability to maximize
More informationInfluence of Human Resource Management Practices on the Performance of Small and Medium Enterprises in Kisumu Municipality, Kenya
International Journal of Business and Social Science Vol. 4 No. 1; January 2013 Influence of Human Resource Management Practices on the Performance of Small and Medium Enterprises in Kisumu Municipality,
More informationTesting for oil saving technological changes in ARDL models of demand for oil in G7 and BRICs
Sustainable Development and Planning V 947 Testing for oil saving technological changes in ARDL models of demand for oil in G7 and BRICs M. Asali Petroleum Studies Department, Research Division, OPEC,
More informationWhat STIRPAT tells about effects of population and affluence on environmental impact?
What STIRPAT tells about effects of population and affluence on environmental impact? Taoyuan Wei 1 July 21, 2010 Abstract In the literature of STIRPAT application to environmental impacts of population
More informationAssessment of the Contributions of Bee-keeping Extension Society to the Income of Bee-Farmers in Kaduna State
PAT 2008; 4 (1): 28-37: ISSN: 0794-5213 Online copy available at www.patnsukjournal.com/currentissue Assessment of the Contributions of Bee-keeping Extension Society to the Income of Bee-Farmers in Kaduna
More informationStatistical Observations on Mass Appraisal. by Josh Myers Josh Myers Valuation Solutions, LLC.
Statistical Observations on Mass Appraisal by Josh Myers Josh Myers Valuation Solutions, LLC. About Josh Josh Myers is an independent CAMA consultant and owner of Josh Myers Valuation Solutions, LLC. Josh
More informationA Parametric Bootstrapping Approach to Forecast Intermittent Demand
Proceedings of the 2008 Industrial Engineering Research Conference J. Fowler and S. Mason, eds. A Parametric Bootstrapping Approach to Forecast Intermittent Demand Vijith Varghese, Manuel Rossetti Department
More informationExport Growth and Poverty Reduction in Nigeria
IOSR Journal Of Humanities And Social Science (IOSR-JHSS) Volume 16, Issue 3 (Sep. - Oct. 2013), PP 48-52 e-issn: 2279-0837, p-issn: 2279-0845. www.iosrjournals.org Export Growth and Poverty Reduction
More informationHotel Industry Demand Curves
Journal of Hospitality Financial Management The Professional Refereed Journal of the Association of Hospitality Financial Management Educators Volume 20 Issue 1 Article 6 Summer 2012 Hotel Industry Demand
More informationFORECASTING THE GROWTH OF IMPORTS IN KENYA USING ECONOMETRIC MODELS
FORECASTING THE GROWTH OF IMPORTS IN KENYA USING ECONOMETRIC MODELS Eric Ondimu Monayo, Administrative Assistant, Kisii University, Kitale Campus Alex K. Matiy, Postgraduate Student, Moi University Edwin
More informationFAO Global Database on Agricultural Capital Stock (ACS) and path toward a FAO Agricultural Productivity Database
FAO Global Database on Agricultural Capital Stock (ACS) and path toward a FAO Agricultural Productivity Database Marie Vander Donckt, Statistics Division, FAO FIRST MEETING OF THE OECD NETWORK ON AGRICULTURAL
More informationTRAVEL COST METHOD (TCM)
TRAVEL COST METHOD (TCM) Learning Outcomes Explain the concept of travel cost method Prepare questionnaire to be used in TCM Apply TCM to calculate recreation benefits Introduction TCM is used to value
More informationTable D.1: Time Spent Collecting Wood (hours/day)
Fuelwood Collection and Cooking Table D.1 shows that the time spent collecting wood each day varies quite considerably. But the average time is more than 2.5 hours, and it can be eight times that. The
More informationGlobal Poverty: Recent Trends and Prospects for the Future Will We Achieve to Eradicate Extreme Poverty by 2030?
Global Poverty: Recent Trends and Prospects for the Future Will We Achieve to Eradicate Extreme Poverty by 2030? Prof. Dr. Michael Grimm University of Passau, Erasmus University Rotterdam, RWI Essen, IZA
More informationThe Effect of Minimum Wages on Employment: A Factor Model Approach
The Effect of Minimum Wages on Employment: A Factor Model Approach Evan Totty April 29, 2015 Abstract This paper resolves issues in the minimum wage-employment debate by using factor model econometric
More informationForecasting with Many Predictors: allowing for non-linearity
Forecasting with Many Predictors: allowing for non-linearity Eran Raviv 1,2,* and Dick van Dijk 1,2,3 1 Econometric Institute, Erasmus University Rotterdam 2 Tinbergen Institute 3 Erasmus Research Institute
More informationLow-quality, low-trust and lowadoption: Saharan Africa. Jakob Svensson IIES, Stockholm University
Low-quality, low-trust and lowadoption: Agriculture in Sub- Saharan Africa Jakob Svensson IIES, Stockholm University This talk Technology adoption in agriculture Use (or rather none-use) of fertilizer
More informationA Statistical Analysis on Instability and Seasonal Component in the Price Series of Major Domestic Groundnut Markets in India
International Journal of Current Microbiology and Applied Sciences ISSN: 2319-776 Volume 6 Number 11 (217) pp. 815-823 Journal homepage: http://www.ijcmas.com Original Research Article https://doi.org/1.2546/ijcmas.217.611.96
More informationChapter 3. Table of Contents. Introduction. Empirical Methods for Demand Analysis
Chapter 3 Empirical Methods for Demand Analysis Table of Contents 3.1 Elasticity 3.2 Regression Analysis 3.3 Properties & Significance of Coefficients 3.4 Regression Specification 3.5 Forecasting 3-2 Introduction
More informationTo combine or not to combine: selecting among forecasts and their combinations
International Journal of Forecasting 21 (2005) 15 24 www.elsevier.com/locate/ijforecast To combine or not to combine: selecting among forecasts and their combinations Michèle Hibon*, Theodoros Evgeniou
More informationMicrofinance increases latrine uptake in rural Cambodia Results from a randomized controlled trial
Microfinance increases latrine uptake in rural Cambodia Results from a randomized controlled trial Tamara Baker ide Global WASH Initiative OU WaTER Conference September 23, 2013 Agenda 1. Background on
More informationImpacts of Farmer Inputs Support Program on Beneficiaries in Gwembe District of Zambia
Impacts of Farmer Inputs Support Program on Beneficiaries in Gwembe District of Zambia Alfred Sianjase Ministry of Agriculture and Livestock Government of Zambia E-mail: alsianjase@gmail.com Venkatesh
More informationCREDIT RISK MODELLING Using SAS
Basic Modelling Concepts Advance Credit Risk Model Development Scorecard Model Development Credit Risk Regulatory Guidelines 70 HOURS Practical Learning Live Online Classroom Weekends DexLab Certified
More informationThe Impact of Kinship Networks on the Adoption of Risk-Mitigating Strategies in Ethiopia
The Impact of Kinship Networks on the Adoption of Risk-Mitigating Strategies in Ethiopia Salvatore Di Falco and Erwin Bulte José Victor Cremonesi Giarola Carlos Monge-Badilla Université Paris 1 Panthéon-Sorbonne
More informationActive Learning for Logistic Regression
Active Learning for Logistic Regression Andrew I. Schein The University of Pennsylvania Department of Computer and Information Science Philadelphia, PA 19104-6389 USA ais@cis.upenn.edu April 21, 2005 What
More informationA Treeboost Model for Software Effort Estimation Based on Use Case Points
Western University Scholarship@Western Electrical and Computer Engineering Publications Electrical and Computer Engineering 12-2012 A Treeboost Model for Software Effort Estimation Based on Use Case Points
More information